Best of Technical Support

I'm trying to upgrade to kernel 2.2.5. When I type mkinitrd /boot/initrd-2.2.5.img 2.2.5, the system returns:

mount: the kernel does not recognize /dev/loop0 as a block device
Can't get a loopback device

What should I do? What has gone wrong? —Mun Hon Tham, mhtham@cyberway.com.sg

For some reason, automatic module loading is broken (you could do modprobe loop to load the module by hand). I am willing to bet you installed all the updates from the Red Hat FTP site, including the new modutils from the 2.2 directory. Unfortunately, modutils has a completely broken kerneld that causes the message you are seeing. I do not know why Red Hat hasn't taken that broken package off their FTP site. In the meantime, all you need to do is downgrade to the previous modutils by typing:

rpm -U --force modutils-version.i386.rpm

—Marc Merlin, marc@merlins.org

The Installation guide says the partition that contains /boot must be below cylinder 1023. My hard drive has 1400 cylinders. How do I find out which cylinders are not being used and if /boot will fit below cylinder 1023?

I will be using a dual-boot system with a DOS partition. My hard drive is 10.9GB and DOS/Windows is the primary partition, using 4.5GB. Can you help? —Jason Hipsher, Jclive@netscape.net

The most important factor that slips up new users is LBA mode drives. Logical Block Addressing translates cylinders into heads, dividing one by two and multiplying the other by two, so the numbers still work but you have fewer cylinders.

The problem is that with a 10GB drive, you most likely do not have 1400 cylinders. You probably have far more, and LBA is enabled on your drive (this is done in your BIOS and is normally enabled by default). You cannot turn it off without reloading your DOS/Windows partition.

You may be able to use the “linear” addressing feature in LILO to get around this problem, but it doesn't always work. A neat trick I use to get around the issue is to create a small (8MB is usually more than enough) partition at the very start of my drive to store my kernels. You may need PartitionMagic or some other partition-manipulation tool to do this without loss of data, but it will give you a place to store files that is guaranteed to be lower than the 1024th cylinder. You can then mount this at boot time (I put it under /img), so you can store your kernels there. Linux itself can be anywhere on the drive. —Chad Robinson, chadr@brt.com

Using the installation's fdisk (or disk druid), create the partitions you will need with the first one for DOS/Windows. Then separate a small (64-128MB) partition for swap and the rest for Linux. When you finish your installation, make sure you pick “MBR install” for LILO. —Mario Bittencourt, mneto@argo.com.br

What, if any, is the difference between X11 and X Windows? It seems as though both terms can be used interchangeably, but at the same time, it seems as if they are different servers. Is X Windows a shell or server on top of X11, or are they entirely different? I am not talking about look and feel, but the basic meaning. —Chris, kender_kin@hotmail.com

The terms are often used interchangeably, which can be confusing. The “official” name for X is “The X Window System”. Some purists get irritated when people say “X Windows”, but it's usually safe to say “X”.

You are actually talking about the same program. What's different is that the first term (X11) refers to the specific version of X. You're actually missing part of it—an Rx is usually included after it. For example, X11R5 refers to a popular and still commonly seen release. X11R6 is what most new Linux distributions are shipping with now (in the form of XFree86, an open-source distribution of X). —Chad Robinson, chadr@brt.com

When I type startx, the response I get is:

X11TransSocketUNIXConnect : Can't connect: errno = 111
giving up
xinit: Connection refused(errno 111) unable to connect to X server
xinit: no such process(errno 3) server error

Now, this is kind of irritating—I have used XF86config, but it doesn't seem to have the driver for the SIS6215 (4MB RAM). We have had X up, but only after we made some adjustments—we could view it in only 16 colors. Since this is just the 2.0.36 kernel, is there a chance that I must upgrade it? Do I need to get the XFree3.3.3 version to get this thing alive? (The school's hardware is a P166, SIS6215 videocard, Digital screen.) Do I have to reconfigure everything, or maybe install the new kernel and XFree version? —Mathias Bakkejord, chenrazee@hotmail.com

The newer the X version you use, the better your chances of finding a driver designed for your card. Upgrading your Linux kernel won't help here, although there are other good reasons to do so.

However, you can probably get your card working in at least 256-color mode by using the SVGA driver. This is a good idea anyway, since it will give you a chance to learn what the clock rate and refresh settings for your card and monitor need to be. It would be a good idea to try this before trying to find a specific driver for your card. —Chad Robinson, chadr@brt.com

I would like to change the pre-login banner on my Red Hat 5.1 system; however, each time I change /etc/issue, “the system” changes it back. How do I get around this? —Philip Lowe, uplowe@mcs.drexel.edu

The banner file /etc/issue is created by the script file /etc/rc.d/rc.local which runs whenever the machine is booted. In order to modify the issue file, you'll need to edit the relevant lines in rc.local; either comment out the lines that overwrite the current issue file, or incorporate your changes so the script generates the file you are after. The lines in question are documented with comments, so you should have no problem finding the lines to change. —Vince Waldon, Vince.Waldon@cnpl.enbridge.com

I am new to Linux. I have successfully installed Caldera OpenLinux 2.2 and have configured KDE, etc. I want to use kppp to dial up to the Internet. However, when I attempt to dial up, I get an error message indicating “Sorry, the modem is busy.”

I have run LISA to install the modem on /dev/ttys1 (Com Port 2). I have tried setting kppp to /dev/modem/ and several other tty devices (when set to ttys0, I get a different message indicating the modem cannot be found). The modem is an internal US Robotics, 56K modem (not a winmodem). All attempts to access the modem (e.g., via Test Modem in the kppp setup) result in the same message.

I have referenced several texts, including the kppp Handbook and followed all the tips; but the modem still has the same problem. I have researched the Internet newsgroups where several other people have mentioned this issue, but no one seemed able to answer the question. —Shannon Brown, linux@rubicongulch.com

Make sure you have no lock file for your modem. Your first attempts to set it up may have accidentally terminated a process before it could remove its lock file. You normally find these in /var/spool/locks, depending on the program that creates the lock. You should probably use /dev/ttyS1 (note the upper case S) to access your modem if it is on COM2 (or /dev/ttyS0 for COM1). —Chad Robinson, chadr@brt.com

I don't know how to use PCNFSD to enable lockd and share. Red Hat technical support refuses to answer me. Where can I get documentation and how can I implement lockd? —Joan Cartigny, cartigny@bourely.com

Last I checked, the tech support you get with a Red Hat box is installation tech support, not detailed configuration. If you need configuration tech support, you need to purchase a support contract with Red Hat, LinuxCare or another provider.

Are you sure you mean PCNFSD which is used only by some DOS/Windows clients? If you mean NFS, you need to use knfsd to get lockd support, and for this, you should use a 2.2.x kernel. You can upgrade yourself, or get Red Hat 6.0 which includes them by default. —Marc Merlin, marc@merlins.org

I'm trying to install KDE. When I try to run it, it says I'm missing libstdc++2.9. I asked about this on a mailing list and was told that I was probably missing a C compile library. I then added every C compiler listed in disk d1 to my Slackware 3.6 installation. Still didn't have libstdc++2.9. On an official KDE site, it says I'm supposed to edit the .xinitrc file, but when I attempt to do that in vi, it says .xinitrc is not a regular file and then declines to edit it.

Any and all information about the Slackware 3.6 installation of KDE would be helpful. —Lisa Zuckerman, blueink@netzero.net

This is bad news because .xinitrc is a regular text file—perhaps yours is corrupted.

Slackware is missing many recent libraries and is still using libc5. If you grabbed KDE binaries, they are most likely compiled against glibc, and they won't work on your machine. libstdc++ is a C++ library that comes with glibc, I believe.

You can find and install all the libraries yourself, but I recommend switching distributions. The most recent ones have KDE built-in. Slackware is usually a year behind, although a more recent version of Slackware may help. —Marc Merlin, marc@merlins.org

I have MkLinux installed on my Macintosh notebook computer (PB 3400/200). When I am at work, I like to use my Sun (Ultra 2) to display Windows from my powerbook. To do this, I use telnet to get to my powerbook from the Sun:

setenv DISPLAY SunComputerName

At the Sun prompt, I type:

xhost + MacName

You are apparently missing fonts on the Sun machine. I believe Applixware has some documentation on how to run the application remotely or over NFS. The idea is that you must install the required fonts on your Sun. LyX probably has the same problem. —Marc Merlin, marc@merlins.org